Water resource management in agriculture

• Cultivator has to book for water one year in advance
• Administration has to make sure there is twice the quantity which has been booked for the whole region.
• Domain expert (INRA) wants to defines and assess new cultures activities

Multiple stakeholders, multiple concerns and scales

Decisions

• Which field to use for growing what ?
• How would the crops grows?
• When would I have to add water ?
• Machines ? Peoples ?

Complex System, optimization with multiple factors (weather, costs, resources)

How our technologies and techniques would fare in such context?

Disclaimer: this experiment is not about the science itself but about how to use the OSS modeling technologies.
This is a toy, but a toy complex enough that we can learn from it.

Process

1. State your resources (Machines, peoples, fields)
2. Pick some climate model
3. Assign surfaces to cultures
4. Deduce a possible planning
5. See how the biomass would grow
6. See how much irrigation we would need
7. do it again !

Demo time

EMF

Several Ecore models

👍 API

						
 public static void main(String[] args) {
    // ...
    Exploitation exploitation = loadFromFile(args[1]);
    for (WorkGroup group : exploitation.getGroups()) {
      System.out.println(group.getName());

      for (Culture cultureAssignedToThisGroup : group.getCultures()) {
        for (ExploitationActivity activity : cultureAssignedToThisGroup
            .getActivities()) {
          System.out.println(activity.getName());
          System.out.println(activity.getFrequency());
          System.out.println(activity.getStartDate());
          System.out.println(activity.getEndDate());
        }
    // ...
					
					

👍 Seamless data reuse

						
 public class ClimateDataResourceImpl extends ResourceImpl {

  public ClimateDataResourceImpl(URI uri) {
    super(uri);
  }

  @Override
  protected void doLoad(InputStream inputStream, Map options)
      throws IOException {
    getContents().clear();
    ClimateData climate = SimulationFactory.eINSTANCE.createClimateData();
    String content = CharStreams.toString(new InputStreamReader(inputStream,
        Charsets.US_ASCII));

    List lines = Lists.newArrayList(Splitter.on('\n')
        .omitEmptyStrings().split(content));

    // ...
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
						
					

Textual Syntax : Xtext

👍 predicates, expressions, conditions

Graphical Syntax: Sirius

👍 Dynamic Editing

👍 Table editors

👍 Animation*

Barely scratched the surface

• Eclipse ICE and Science WG
• Generating reports using Acceleo
• Comparing alternatives using EMF Compare
• ...

Analysis

Planning problems

If you need to optimize goals under constraints while having limited resources.

When will I seed the crop based on the culture constraints, the climate and the peoples, machine, fields I have ?

NP class of problems, huge number of solutions, many being useless

👍 Easy to mix with EMF

						
@PlanningEntity(
     difficultyComparatorClass = ActivityWorkDifficultyComparator.class
     )
public interface ActivityWork extends EObject {
//...
@PlanningVariable(valueRangeProviderRefs = {"days"})
	Day getScheduledOn();
//...	
}
					

						
@PlanningEntity(
     difficultyComparatorClass = ResourceAllocationDifficultyComparator.class
     )
public interface ResourceAllocation extends EObject {
//...
@PlanningVariable(valueRangeProviderRefs = {"resources"})
	Resource getResource();
//...	
}
					

👍 Easy to mix with EMF

						
@PlanningEntity(
     difficultyComparatorClass = ActivityWorkDifficultyComparator.class
     )
public interface ActivityWork extends EObject {
//...
@PlanningVariable(valueRangeProviderRefs = {"days"})
	Day getScheduledOn();
//...	
}
					

						
@PlanningEntity(
     difficultyComparatorClass = ResourceAllocationDifficultyComparator.class
     )
public interface ResourceAllocation extends EObject {
//...
@PlanningVariable(valueRangeProviderRefs = {"resources"})
	Resource getResource();
//...	
}
					

👍 Easy to mix with EMF

						
@PlanningEntity(
     difficultyComparatorClass = ActivityWorkDifficultyComparator.class
     )
public interface ActivityWork extends EObject {
//...
@PlanningVariable(valueRangeProviderRefs = {"days"})
	Day getScheduledOn();
//...	
}
					

						
@PlanningEntity(
     difficultyComparatorClass = ResourceAllocationDifficultyComparator.class
     )
public interface ResourceAllocation extends EObject {
//...
@PlanningVariable(valueRangeProviderRefs = {"resources"})
	Resource getResource();
//...	
}
					

Score Function

						
public Score calculateScore(SimulationSolution solution) {
  int hardScore = 0;
  int softScore = 0;
  //...
  /*
   * Constraint: all the resources which are required should be allocated.
   */
  for (ResourceAllocation alloc : solution.getSimulation().getAllocations()) {
    if (alloc.getResource() == null) {
      hardScore += mediumPenalty(1);
      addFeedback(
          alloc.getWork(),
          (createFeedback(FeedbackLevel.ERROR, "A required resource of kind "
              + alloc.getKind().getName() + " is missing.")));
    }
  }
  
  //...
 return HardSoftScore.valueOf(hardScore, softScore)
 }

					

What have we learned so far?

“Do not compromise on your domain model.”

Effort to build this tooling ?

• 3 hours meeting + mail exchanges with INRA experts
• 10 days of Eclipse Modeling experts
• Proof of concept code is on github

Technologies meant to do this

Are relevant to save this

Teaming up technologies instead of stacking it

Eclipse: platform and User interface integration

EMF: data, resources, deeplinkink, reflective manipulation

Opportunities

• Time modeling and management
• Probabilistic models
• Graph/Charts representation in Sirius
• Alternatives comparisons
• ...

Further Materials